You are on page 1of 4



Mark Hyman,
Joseph Pizzorno, , Andrew Weil,

Interpreting Studies: The Importance of Context and the balance is protective, and if so, how can we best achieve that goal?
Role of the Scientist Nutrients, in general, do not lend themselves to a research design
The recent meta-analysis1 and media coverage surrounding that is best suited for pharmacological interventions because,
the risks and benefits of vitamin E repeat a common failure of unlike pharmaceuticals, nutrients are multi-functional substances
physicians and mediathe sensationalization of medical built into our evolutionary design. There are many confounding fac-
research reports combined with the lack of context and nuance tors in any randomized trial and meta-analysis. The vagaries in indi-
that leaves practicing physicians, patients and consumers con- vidual diets, habits, pre-clinical and clinical disease, genomic
fused. Fortunately a more dispassionate and rational view plac- diversity of SNPs (single nucleotide polymorphisms), the bioavail-
ing the current meta-analysis in the context of a broader ability and biological specificity of different forms of individual nutri-
literature on vitamin E and antioxidants can be a better guide to ents, and selection of the most effective combinations of nutrients,
recommendations and policy. make nutritional research through randomized trials and meta-
Medical science is imprecise and the tools of analysis and analyses very problematic. Given that a methodologically sound,
research are imperfect despite our best intentions. No statistical definitive randomized placebo-controlled trial has not been done and
model or research methodology can accurately predict the outcome may never be done, the best we can currently do is take the best avail-
in all humans who vary greatly in habits and genes. Medicine has in able evidence from basic science, studies of redox status, nutrient lev-
recent years been humbled in its hunger to help by its mistakes els, trials with clearly measurable intermediate endpoints, relevant
including hormone replacement therapy,2 which in the end hurt biomarkers, gene expression analysis, along with observational data
rather than helped the heart, the lipids only hypothesis of cardio- and large-scale randomized trials and then synthesize all the points of
vascular disease that ignores the important role of inflammation,3 evidence into a biologically plausible theory, clinical guidelines and
and the belated,4 sober end to our love affair with COX-2 inhibitors. when justified public policy. This was not done.
Uncertainty is common in medicine. In the fact of uncertainty, Meta-analysis confronts us with many methodological diffi-
we must take a broader view, step back from the canvas of individ- culties including publication bias; heterogeneity of the trials; varia-
ual studies and meta-analyses and view the landscape of medicine tions in quality, design and purpose of the individual trials, as well
and scientific inquiry as a whole, while understanding the limita- as the fact that the findings are often not borne out in later random-
tions and value of different tools of research. We require all the ized trials on the same subject.5 LeLorier admonishes us that the
tools in order create sense and be sensible in a shifting sea of data popularity of meta-analysis may at least partly come from the fact
points. The goal of seeking an informed balanced overview is to that it makes life simpler and easier for reviewers as well as readers.
help us create understanding from knowledge, to sort through the However, oversimplification may lead to inappropriate conclu-
facts and place them in the context of biologic principles, against sions. Furthermore, conclusions of many studies noted in
the backdrop of all we know. Then we ask ourselves if the new data abstracts of major journals are at odds with data published in the
fit into the landscape or appear like a polar bear in a desert. We same paper.6 This oversimplification and conclusions inconsistent
must also recognize that the questions we ask, how we ask them, with the data were evident in the recent-metanalysis on vitamin E.
and why, all inform the answers we receive.
The responsibility of the scientist is to filter the research, place Reviewing the Review: Errors in Design and Interpretation
it in context, provide hope where appropriate and caution where With these caveats in mind, a brief examination of the basic
necessary, but most importantly to be an informed, measured science, observational data, randomized trials and interventional
guide to a public seeking to gain health and ameliorate suffering. studies with clear intermediate endpoints can provide a more ratio-
What can we learn from the recent example of vitamin E? There are nal context and framework for recommendations about vitamin E,7
a number of limitations from the recent vitamin E meta-analysis as well as a model for assessing the role of nutrients and other bio-
and opportunities to refocus our research strategy. Using long-term logical therapies in health and disease.
randomized trials and meta-analyses for the study of lifestyle inter- The data in this meta-analysis pooled 135,967 subjects from
ventions, diet and individual nutrients is fraught with difficulty in 19 studies, which individually did not show any statistically signifi-
design and effective separation of the plethora of their biologic cant increase in mortality. The 19 studies were selected from 2,170
effects. We may, in fact, be asking the wrong question. Asking if studies. The authors concluded, high dose (>400 IU/d) vitamin E
high dose vitamin E is helpful or harmful may not provide the supplements may increase all-cause mortality and should be avoid-
information we seek. We may better ask if restoring normal redox ed. However the authors in the body of the paper provide ade-

14 ALTERNATIVE THERAPIES, jan/feb 2005, VOL. 11, NO. 1 A Rationale Approach to Antioxidant Therapy and Vitamin E
quate caution about over interpretation of the results. They proper- years of use of these vitamins. Any benefit might have been miti-
ly raise concerns related to publication bias; the small size of most gated by the use of the potentially harmful synthetic beta-carotene,
of the included trials; the inclusion of patients with chronic disease and the synthetic racemic-alpha-tocopherol. While the study did
limiting relevance to healthy adults; the concomitant use of other measure serum levels of antioxidants, redox status was not
vitamins including beta-carotene that has shown an increase in assessed, leaving open the question of whether the anti-oxidants
morality in smokers and drinkers; the lack of homogeneity of toco- achieved their expected biologic effects.
pherol isomers among studies (inconsistent use of synthetic all-rac The AREDS trials are multiple and the selection of the nega-
alpha tocopherol or naturally occurring RRR-alpha-tocopherol); tive trials for inclusion in the analysis is perplexing despite publica-
and the non-statistically significant reduction in mortality from tion of conflicting results in the same issue of the Archives of
doses less than 200 IU. The benefits for intermediate endpoints in Ophthalmology. Report No. 9,12 included in the meta-analysis,
many of the trials cited (Cambridge Heart Antioxidant Study showed no benefit in reduction of cataracts from high dose com-
[CHAOS],8 SPACE9) such as the reductions in nonfatal myocardial bined anti-oxidant therapy with beta-carotene 15mg, vitamin C
infarction (MI) by 70% to 77% were not discussed. These limita- 500mg, and vitamin E 400 IU either alone or combined with high-
tions were not included in media coverage, nor taken into consider- dose zinc 80mg or copper 2mg (using an insoluble, non-absorbed
ation by many of the expert commentators on the results. form cupric oxide). This is not unexpected because the primary
Other shortcomings of the meta-analysis by Miller et al, are mechanism of cataract formation appears to be glycation, not oxi-
highlighted by a more careful review of the three flawed larger trials dation. Yet an analysis of different endpoints from the same data in
on which the conclusion that high dose vitamin E supplements have Report No. 814 showed a 34% reduction in age-related oxidative
a relative risk of death of 1.04 is based. These were the MRC/BHF11 stress mediated macular degeneration in 3,640 enrolled study par-
Heart Protection Study, CHAOS,8 and AREDS12 (Age-Related Eye ticipants, aged 55-80 years, over 6.3 years. No statistically signifi-
Disease Study) Report No. 9, which represented two-thirds of the cant serious adverse effect was associated with any of the
participants in the high-dose trials. The data of the other trials is less formulations in either analysis. Report No. 1115 estimated that if all
than convincing, and the interpretation and selective omission of persons at risk for AMD (8 million) were to receive the antioxidants
key findings of the CHAOS, AREDS and MRC/BHF trials deserve in the AREDS trial, the 300,000 of the 1.3 million at high risk for
closer scrutiny because they serve as the primary justification for the advanced AMD would avoid that fate and provide considerable
conclusions that vitamin E increases mortality. Of the 19 trials, 11 public health benefits.
showed reduction in mortality, including 7/8 low-dose and 4/11 The studies cited either showed benefit in non-fatal interme-
high-dose trials (which comprised only 30% of patients in the entire diate endpoints such as MI, macular degeneration, arteriosclero-
meta-analysis). It is also important to note that the all cause mortali- sis, angina, prostate and colon cancer, or at the very least showed
ty was available, except for one study, with combined high and low no harm. The conclusions in many studies do not accurately
doses and that for the purposes of the meta-analysis, they doses reflect the data. In the WAVE16 trial of hormone replacement and
were averaged. In addition, the statistical significance of the data is antioxidant interventions in post-menopausal women with coro-
marginal. In the 11 high dose trials, the risk difference was 39 per nary artery disease, the antioxidant treatment group was sicker,
10,000 persons, but the confidence interval was only 3 to 74 per had more smokers and hypertensives, and less patients taking pro-
10,000 persons and the risk ratio was 1.04 with a confidence interval tective treatments with angiotensin converting enzyme inhibitors
of only 1.01 to 1.07 indicating that the risk may have been close to or aspirin, so it is not surprising that the vitamin treatment group
zero, even in the high dose trials. showed a statistically non-significant trend toward worse cardio-
The CHAOS investigators assessed the effect of RRR-alpha- vascular outcomes. The authors found that neither the hormone
tocopherol at 400 or 800 I.U. in 1,035 patients with ischemic heart nor antioxidant group showed benefit and suggested harm from
disease (967 more received a placebo) for 510 days. The vitamin E both therapies as noted in the conclusion; however, the lead
supplementation reduced the risk of non-fatal MI by 77%. There was author of the study, in a reply to a letter to the editor, conceded
a small statistically non-significant increase in cardiovascular deaths that, these results do not prove that high-dose vitamin C and E
in the treatment group; however, those patients were sicker at base- supplements are harmful.17
line, the deaths occurred very early in the study, and further analysis The meta-analysis did not put its results in perspective by
revealed that those patients were actually non-compliant with the reviewing the context of research on vitamin E including the many
vitamin E. Later mortality estimates from CHAOS came from a positive observational,18,19 and interventional studies,20 a large body
research letter, not a peer-reviewed study, 18 and included data after of basic science research and careful analysis of endpoints other
the study was officially ended, and thus subject to information bias than total mortality from the studies in the meta-analysis. It is
due to lack of complete reporting. important to note that many of the larger randomized trials
The single largest trial, the MRC/BHF Heart Protection Study, showed statistically insignificant or barely significant benefit or
of 20,536 high-risk individuals using synthetic or racemic-alpha- harm. Also randomized controlled trials on vitamin E and cardio-
tocopherol of 600mg a day combined with 250mg of vitamin C and vascular disease were subjected to meta-analysis in three other
20mg of beta-carotene (known to increase cancer risk in high-risk studies;21-24 and no significant increase in cardiac or all cause mortal-
individuals). The authors concluded that the results effectively ity was found with doses of up to 800 IU per day. If vitamin E or the
ruled out any substantial reductions, or, indeed, increases, in heart antioxidant theory is useful in preventing and treating disease why
attacks, strokes, cancers, or other major adverse events during 5 do we not see a larger effect? Is it that this question cannot be

A Rationale Approach to Antioxidant Therapy and Vitamin E ALTERNATIVE THERAPIES, jan/feb 2005, VOL. 11, NO. 1 15
answered through randomized trials because of the enormous tocopherols (racemic and RRR, synthetic and natural) relevant
methodological challenge, or is it that we must revisit the proposed when diets contain mostly gamma tocopherols, and the frequently
value of anti-oxidants in secondary prevention? Any new study used racemic form containing l-alpha tocopherol may inactivate
should be considered in the context of the best available evidence RRR-alpha-tocopherol? How do the other forms of antioxidants
from animal studies, basic science, and observational and small and nutrients used in the studies affect the redox balance, signaling
and large interventional trials. and gene expression that determine harm or benefit? How do gene
polymorphisms play a role in nutrient effects? Did the high-fre-
The Role of Vitamin E in Health and Disease quency polymorphism of endolethial nitric oxide synthase (eNOS)
The data on vitamin E collectively point to potentially impor- in the English patients in the CHAOS trial lead to the 77% reduction
tant beneficial biologic effects that are at odds with the reported in non-fatal MI with vitamin E by improving endothelial function?
mortality data. While the basic science must also be cautiously Did differences in pre-treatment nutrient levels because of geo-
used in making clinical assumptions, there are many reasons to graphic location, dietary preferences and nutritional status impact
hypothesize a useful role of vitamin E in promoting health as well outcomes? Would patient stratification by clinical profile and
as preventing and treating disease. A recent review on vitamin E assessment of biomarkers of vitamin E action including effects on
outlined evidence that inhibits LDL oxidation, acts on coagula- redox status, endothelial function, immune function, cell signaling
tion, platelet aggregation and endothelial aggregation, down-regu- and gene expression have provided insight on the possible risks and
lates cellular adhesion molecules, inhibits release of IL-1 beta from benefits with different population groups?
lipopolysaccharide activated monocytes, inhibits protein kinase C New data is best interpreted in the light of existing recom-
affecting a broad array of cell signaling molecules and reduces mendations and findings of independent and governmental bod-
inflammation and smooth muscle cell proliferation, induces apop- ies designed to guide policy. The recommendations from the
tosis, and enhances cell mediated immunity.21 Non-antioxidant Institute of Medicine,31 the US Preventive Services Task Force,32 the
activities of vitamin E include its influence on the activity of several Lewin Group,33 and literature reviews published in the Journal of
other enzymes (eg, PP2A, COX-2, 5-lipooxygenase, nitric oxide the American Medical Association34 and the New England Journal of
synthase, NADPH-oxidase, superoxide dismutase, phopholipase Medicine35 on vitamin therapy all contradict the findings of this
A2) and it also modulates the expression of genes that are involved meta-analysis. In fact, an independent analysis of the data on the
in atherosclerosis (eg, scavenger receptors, integrins, selectins, clinical and economic impact of vitamins in health disease by the
cytokines, cyclins).25 There are clinical trials indicating that vita- Lewin Group for Wyeth Consumer Healthcare found that the
min E normalizes autonomic tone in diabetics,26 reduces risk for potential savings from multi-vitamin use in the Medicare popula-
macular degeneration,12 prostate, and colon cancer,27 reduces MI, tion was $3.9 billion over 5 years from a reduction in cardiovascu-
retards arteriosclerosis,26 and slows progression of neurodegenera- lar and improved immune function and reduction in colon caner
tive disease27 as well as showing benefits in intermittent claudica- with no attributable risk.
tion, fibrocystic breast disease, premenstrual syndrome, In the light of this complex set of data it is inappropriate for
osteoarthritis, and infertility.30 Because vitamin E is actually a mix- the media to publicize isolated findings of selected studies and
ture of eight compounds (alpha, beta, gamma and delta toco- propagate premature conclusions.36 It is even more concerning
pherols and alpha, beta, gamma and delta tocotrienols), and when the academic or practicing physician does not bring an
because they are an integral part of cell signaling mechanisms, and appropriately open mind to consideration of complex biologic
the integrated redox system including ascorbate, carotenoids, r- phenomena, namely oxidative stress in health and disease, and
lipoic acid, and glutathione, the ideal dosage, form and blend of thereby avoid measured, thoughtful and dispassionate discussion
vitamin E and other antioxidants and cell signaling molecules is of any and all interventions related to this possible risk factor,
needed to clearly determine benefit or harm. whether natural or pharmacologic. Our duty is to educate, inform
and provide guidance, not reflexively react to the shifting sands of
Responsibility in Science and the Media scientific discovery or fail to understand the importance of consid-
Unfortunately answering these questions through randomized ering new findings in the context of all levels of data and applying
clinical trials may not be feasible because of the cost, time and multi- the rule of reasonableness.
ple variables required. We cannot assume that testing all nutrients
as drugs in randomized clinical trials even using a single agent with Mark Hyman, MD
a single variable will be feasible or if done will provide us with useful Editor-in-Chief
information with adequate external validity. Understanding the Alternative Therapies in Health and Medicine
role of nutrients as part of the complex array of biologic response
modifiers found in our diet and our environment should eventually Joseph Pizzorno, ND
bring more clarity and guidance for recommendations. Editor
Employing the linear reductionist model using varying study Integrative Medicine: A Clinicians Journal
designs raises many questions regarding the validity of the data.
Are the data in a population of patients with pre-existing cardiovas- Andrew Weil, MD
cular, diabetes and other chronic illness applicable to a healthy Program for Integrative Medicine
population interested in prevention? Are results based only alpha University of Arizona

16 ALTERNATIVE THERAPIES, jan/feb 2005, VOL. 11, NO. 1 A Rationale Approach to Antioxidant Therapy and Vitamin E
References 18. Stampfer MJ, Hennekens CH, Manson JE, et al. Vitamin E consumption and the risk of
1. Miller ER III, Pastor-Barriuso R, Dalal D, et al. Meta-Analysis: High-Dosage Vitamin E coronary disease in women. N Engl J Med. 1993 May 20;328(20):1444-1449.
Supplementation May Increase All-Cause Mortality. Ann Intern Med. 2004 Nov 10 19. Rimm EB, Stampfer MJ, Ascherio A, Giovannucci E, Colditz GA, Willett WC. Vitamin E
2. Rossouw JE, Anderson GL, Prentice RL, et al. Writing Group for the Womens Health consumption and the risk of coronary heart disease in men. N Engl J Med. 1993 May
Initiative Investigators. Risks and benefits of estrogen plus progestin in healthy post- 20;328(20):1450-1456.
menopausal women: principal results From the Womens Health Initiative randomized 20. Salonen RM, Nyyssonen K, Kaikkonen J, et al. Antioxidant Supplementation in
controlled trial. JAMA. 2002 Jul 17;288(3):321-333. Atherosclerosis Prevention Study.Six-year effect of combined vitamin C and E supple-
3. Ridker PM. High-sensitivity C-reactive protein, inflammation, and cardiovascular risk: mentation on atherosclerotic progression: the Antioxidant Supplementation in
from concept to clinical practice to clinical benefit. Am Heart J. 2004;148:S19-S26. Atherosclerosis Prevention (ASAP) Study. Circulation. 2003 Feb 25;107(7):947-953.
4. Mukherjee D, Nissen SE, Topol EJ. Risk of cardiovascular events associated with selective 21. Brigelius-Flohe R, Kelly FJ, Salonen JT, et al. The European perspective on vitamin E: cur-
COX-2 inhibitors. JAMA. 2001 Aug 22-29;286(8):954-959. Review. rent knowledge and future research. Am J Clin Nutr. 2002 Oct; 76(4):703-716.
5. LeLorier J, Gregoire G, Benhaddad A, Lapierre J, Derderian F. Discrepancies between meta- 22. Shekelle PG, Morton SC, Jungvig LK, et al. Effect of supplemental vitamin E for the pre-
analyses and subsequent large randomized, controlled trials. N Engl J Med. vention and treatment of cardiovascular disease. J Gen Intern Med. 2004;19(4):380-389.
1997;337(8):536-542. 23. Eidelman RS, Hollar D, Hebert PR, Lamas GA, Hennekens CH. Randomized trials of vit-
6. Pitkin RM, Branagan MA, Burmeister LF. Accuracy of Data in Abstracts of Published amin E in the treatment and prevention of cardiovascular disease. Arch Intern Med.
Research Articles, JAMA. 1999; 281:1110 - 1111. 2004;164(14):1552-1556.
7. Williams R. Biochemical individuality. New York: John Wiley and Sons, 1963. 24. Vivekananthan DP, Penn MS, Sapp SK, Hsu A, Topol EJ. Use of antioxidant vitamins for
8. Neuzil J,Weber C, Kontush A. The role of vitamin E in atherogenesis: linking the chemical, the prevention of cardiovascular disease: meta-analysis of randomised trials. Lancet.
biological and clinical aspects of the disease. Atherosclerosis. 2001;157:257283. 2003;361(9374):2017-2023.
9. Stephens NG, Parsons A, Schofield PM, et al. Randomised controlled trial of vitamin E in 25. Munteanu A, Zingg JM, Azzi A. Anti-atherosclerotic effects of vitamin E--myth or reality?
patients with coronary disease: Cambridge Heart Antioxidant Study (CHAOS). Lancet. J Cell Mol Med. 2004 ;8(1):59-76.
1996;347:781786. 26. Manzella D, Barbieri M, Ragno E, Paolisso G. Chronic administration of pharmacologic
10. Boaz M, Smetana S, Weinstein T, et al. Secondary prevention with antioxidants of cardio- doses of vitamin E improves the cardiac autonomic nervous system in patients with type
2 diabetes. Am J Clin Nutr. 2001 Jun;73(6):1052-1057.
vascular disease in endstage renal disease (SPACE): randomised placebo controlled trial.
27. Virtamo J, Pietinen P, Huttunen JK, et al . ATBC Study Group. Incidence of cancer and
Lancet. 2000; 356: 12131218.
mortality following alpha-tocopherol and beta-carotene supplementation: a postinter-
11. Heart Protection Study Collaborative Group MRC/BHF Heart Protection Study of antioxi- vention follow-up. JAMA. 2003 Jul 23;290(4):476-485.
dant vitamin supplementation in 20536 high-risk individuals: a randomised placebo-con- 28. Fang JC, Kinlay S, Beltrame J, Hikiti H, Wainstein M, Behrendt D, Suh J, Frei B, Mudge
trolled trial. Lancet. 2002; 360: 2333. GH, Selwyn AP, Ganz P. Effect of vitamins C and E on progression of transplant-associat-
12. Age-Related Eye Disease Study Research Group. A randomized, placebo-controlled, clinical ed arteriosclerosis: a randomised trial. Lancet. 2002. 30;359(9312):1108-1113.
trial of high-dose supplementation with vitamins C and E and beta carotene for age-related 29. Sano M, Ernesto C, Thomas RG, et al. A controlled trial of selegiline, alpha-tocopherol,
cataract and vision loss: AREDS report no. 9. Arch Ophthalmol. 2001;119(10):1439-1452. or both as treatment for Alzheimer's disease. N Engl J Med. 1997; 336:1216-1222.
13. Mitchinson MJ, Stephens NG, Parsons A, et al. Mortality in the CHAOS trial. Lancet. 1999 30.
Jan 30;353(9150):381-382. 31.
14. Age-Related Eye Disease Study Research Group. A randomized, placebo-controlled, clinical 32. Morris CD, Carson S. Routine vitamin supplementation to prevent cardiovascular dis-
trial of high-dose supplementation with vitamins C and E, beta carotene, and zinc for age- ease: a summary of the evidence for the U.S. Preventive Services Task Force. Ann Intern
related macular degeneration and vision loss: AREDS report no. 8. Arch Ophthalmol. Med. 2003 Jul 1;139(1):56-70. Review.
2001;119(10):1417-1436. 33. The Lewin Group. A Study of the Cost Effects of Daily Multivitamins for Older Adults,
15. Bressler NM, Bressler SB, Congdon NG, et al. Age-Related Eye Disease Study Research Prepared for Wyeth Consumer Healthcare, January 14, 2004.
Group. Potential public health impact of Age-Related Eye Disease Study results:AREDS 34. Fairfield KM, Fletcher RH. Vitamins for Chronic Disease Prevention in Adults: Scientific
report no. 11. Arch Ophthalmol. 2003;121(11):1621-1624. Review JAMA, Jun 2002; 287:3116 - 3126.
16. Waters DD, Alderman EL, Hsia J, et al. Effects of hormone replacement therapy and 35. Willett WC, Stampfer MJ. What Vitamins Should I be Taking Doctor? N Engl J Med.
antioxidant vitamin supplements on coronary atherosclerosis in postmenopausal 2001; 345:1819-1824.
women: a randomized controlled trial. JAMA. 2002;288:2432-2440. 36. Moynihan, R, Bero, L, Ross-Degnan, D, Coverage by news media of the benefits and risks
17. Waters D. Vitamins and Hormone Therapy for CAD. JAMA. 2003;289:982. of medications. N Engl J Med. 2000 Jun1;342(22):1645-50.

A Rationale Approach to Antioxidant Therapy and Vitamin E ALTERNATIVE THERAPIES, jan/feb 2005, VOL. 11, NO. 1 17